Observation Sensitivity Tests Using the MOM6-Data Assimilation System for the Hurricane Fiona and Ian

Currently, there are two 3DVar data assimilation (3DVar DA) schemes established for the first time for a regional ocean model MOM6. We have conducted data denial experiments for the period of Hurricane Fiona and Ian in 2022. The primary objective is to determine the optimum 3DVar DA configuration for a regional application and investigate the impact of glider observations. The 3DVar DA based on BUMP3D with fixed correlation length scales produces relatively poor temperature (T) and salinity (S) analysis, compared to one based on the explicit diffusion (ED) operator with variable length scales. Through denying sea surface height (SSH), we found that the BUMP3D approach consistently produces poor performances, having relatively large RMSE for T and S. A similar performance exists with glider denial experiments. There were no significant impacts on SST by either SSH or glider denial test. Interestingly, however, we found that the BUMP3D method is associated with less negative skewness in SSH OMA (observations minus analysis), when glider observations are denied. On the contrary, the SSH OMA for the ED operator is opposite. The SST OMA is all positive, but there was no apparent difference in the impact between glider and SSH denial experiments, and also between the two DA methods. The sea surface salinity (SSS) OMA showed negative bias and negative skewness, which was further worse by the BUMP3D method, and both the data denial shows little impact on SSS for both the DA schemes. Overall, we found that the BUMP3D approach underperforms, compared to the explicit diffusion method, showing large RMSE. This might be related to unbalanced SSH in a balance operator. In this talk, we will present the results in more detail, and present a remedy to improve the 3DVar DA schemes.